European Research Studies Journal Volume ΧΧΙΙΙ, Special Issue 2, 2020 pp. 400-409 Forecasting the Level of Unemployment, Inflation and Wages: The Case of Sweden* Submitted 14/09/20, 1st revision 02/10/20, 2nd revision 21/10/20, accepted 17/11/20 Tomasz Grodzicki1, Mateusz Jankiewicz2 Abstract: Purpose: In the macroeconomic theory and analyses there are a number of studies focused on three crucial phenomena, namely unemployment, inflation, and wages. As a result, the term called ‘Phillips curve’ was introduced in order to illustrate a negative correlation between inflation and the unemployment rate. This paper is to establish the relationship between unemployment, inflation, and wages in Sweden, and to forecast their value using. Design/Methodology/Approach: The Vector Autoregression (VAR) model has been used for the analysis. The analysis applies the values of the unemployment rate, the level of the minimum wage and the value of inflation in the period of 2002 - 2017 on a quarterly basis. Findings: Results from the analysis show that (1) the unemployment rate and the level of wages do not explain well enough inflation developments in Sweden and (2) as in many previous empirical studies on this topic, there are no significant changes in employment resulting from the increase in the minimum wage. Practical Implications: This study is of great importance especially for the policy makers who can apply the presented analysis tools to predict further tendencies in shaping value of the main economic indicators and make appropriate decisions to avoid possible recession and its negative consequences. Therefore, the outcomes of this paper can increase efficiency of the economic policy in numerous countries. Originality/Value: This research checks the suitability of the widely known tool to evaluate dependencies and predict further situation of the main economic indicators in Sweden. Moreover, the analysis shows that the real relationships between unemployment rate, minimum wages and inflation are not always the same as these, which appear in the literature. Keywords: VAR modelling, unemployment rate, inflation, minimum wage rate. JEL Code: C01, B22. Paper type: Research article. 1Nicolaus Copernicus University in Torun, Faculty of Economic Sciences and Management, Torun, Poland, e-mail: [email protected] 2Same as in 1, e-mail: [email protected] *Paper presented in ICABE 2020. Tomasz Grodzicki, Mateusz Jankiewicz 401 1. Introduction The relation of unemployment with inflation and the level of wages is one of the main aspects considered in macroeconomic analyses. Phillips (1958) observed a negative correlation between money wage changes and unemployment in Great Britain in the period of 1861-1957. Then, Samuelson and Solow (1960) placed this way of thinking into more generalized form. They made a clear link between inflation and unemployment as follows: when the level of inflation is high, unemployment rate is low, and the other way around. In addition to this, Samuelson was the one who popularised this phenomenon by giving it a term ‘Phillips curve’ (Blaug, 1997; Bochenek, 2016). More than once, in later studies, deviations from his theory were noted. Phelps (1967) in his work proved that this relationship is only short-lived, because the rise in inflation is followed by the increase in nominal wages, which causes their illusory rise. Illusory because its purchasing power does not change. The society, succumbing to the "deceived" increase in wages, takes up a job, causing a drop in unemployment. After some time, when they become convinced that they are wrong about wage increases, unemployment returns to its original level. Moreover, even with rising inflation, employers tend to ignore it and refrain from increasing wages for employees. Wages, and in particular the ratio of the minimum wage to the level of social benefits, are one of the most important determinants of the unemployment level. Studies on changes in the minimum wage indicate the possibility of both an increase and a decrease in employment as a result of its increase (De Fraja, 1999; Neumark et al., 2004). This is related to changes in marginal productivity. According to the classical theory of employment, if an increase in the minimum wage followed by an increase in wages for the remaining group of employees does not cause changes in marginal productivity, then due to higher labour costs, employers dismiss employees whose marginal product value is below the minimum wage (Evans-Pritchard, 1985). However, according to the effective wage theory, under certain conditions, one can even expect an increase in employment (Raff and Summers, 1987; Bradley, 2007). Many studies show a weak relationship between the increase in the minimum wage and the reduction of poverty (Burkhauser and Sabia, 2007; Gindling, 2018). The reason is that in many poor families nobody works so changes in the minimum wage do not matter to them. The prevailing view in research is that changes in the minimum wage level do not affect employment, but only reduce wage inequalities. The aim of the study is to exhibit the relationship between the processes of unemployment, inflation, and wages, based on the example of Sweden, and to forecast their value using the Vector Autoregression (VAR) model. The analysis used the values of the unemployment rate, the level of the minimum wage and the value of inflation (change compared to the previous period) in the period 2002: 01- Forecasting the Level of Unemployment, Inflation and Wages: The Case of Sweden 402 2017: 01 (due to the lack of availability of data on wages in the later period) on a quarterly basis. On the basis of the estimated model, forecasts of the unemployment rate and the minimum wage were made for seven periods ahead (until the end of 2018) and their admissibility was assessed, and in the case of the unemployment rate, also their accuracy. 2. Methodology Three decades ago, Sims (1980) introduced a new macro econometric tool called Vector Autoregression (VAR) model, which is a n-equation, n-variable linear model. In this model, each variable is explained by its own lagged values and in addition it contains present and past values of the remaining n-1 variables. Stock and Watson (2001) described a VAR as a simple framework that “provides a systematic way to capture rich dynamics in multiple time series, and the statistical toolkit that came with VARs was easy to use and interpret”. In this paper, a VAR model is applied in order to determine the linkages between unemployment, inflation, and wages in Sweden, and to forecast their value. In order to determine the relationship between the given processes, the form VAR model was estimated: (1) where: Y1 – minimum wage value, Y2 – inflation value, Y3 – unemployment rate, t – time variable, Q1, Q2, Q3, Q4 – zero-one seasonal variables, ε - random component. For the statistical verification of the model, the Student's t-test was used to verify the significance of the structural parameters of the model, the Jarque-Ber test to check the consistency of the distribution of residuals with the normal distribution, the Ljung-Box test for the presence of autocorrelation in the residual component and the VAR model stability test. The impulse response was also analysed, and the forecasts of the unemployment rate and minimum wages were assessed based on the model results. Tomasz Grodzicki, Mateusz Jankiewicz 403 3. Empirical Results First, the time structure of each of the analysed processes was examined. Preliminary structure analysis showed the presence of an important component of the trend in the case of the minimum wage and the unemployment rate. Seasonality, on the other hand, plays an important role in shaping only the value of wages. In the case of inflation, none of the components of the trend-seasonal structure was significant. As for at least one of the above-mentioned processes, trend-seasonal components are important, they were also included in the estimation of the VAR model. Then, the selection of the delay order of the vector autoregression model, the results of which are presented in Table 1. All information criteria (AIC, BIC and HQC) take the lowest value for the delay equal to one. Therefore, such a delay order was chosen to estimate the VAR model for the formation of the above-mentioned processes in Sweden. Table 1. Values of information criteria for selecting the delay order of the VAR model Time lag loglik p(LR) AIC BIC HQC 1 -257,253 10,613334* 11,505542* 10,956434* 2 -249,722 0,08925 10,66877 11,89556 11,14053 3 -246,856 0,76631 10,90022 12,46158 11,50065 4 -241,417 0,28424 11,03462 12,93056 11,76371 Source: Own calculations using Gretl software. After selecting the delay order, the VAR model was estimated (1). The results of model estimation and verification are presented in Table 2, while Table 3 presents the results of the analysis of the autocorrelation occurrence in the rest of the VAR model equations. The unemployment rate significantly affects the minimum wage process. During the entire period of this study, the increase in the unemployment rate resulted in a decrease in the minimum wage ceteris paribus. Moreover, there are noticeable regularities in the trend-autoregressive structure. Both the time variable and the seasonal component as well as the autoregressive parameter are statistically significant. In the case of the inflation analysis, a significant impact of the minimum wage and the trend-autoregressive structure were noted. In the analysed period, the relationship between inflation and wages is negative. The unemployment rate, on the other hand, depends on the level of inflation and seasonal and autoregressive parameters. Despite the initial observation of the trend in the level of the unemployment rate, the time variable is not statistically significant. In the case of the first and third equations, the determined coefficient of determination R2 indicates a high degree of explanation of the variability of the dependent variable by the model.